What is Transcript chunk?

A transcript chunk is a short span of a video's spoken words, with aligned start/end timestamps, treated as an independently retrievable unit by an AI search engine's video pipeline — the video equivalent of a passage in text passage-level optimization. When a sub-query matches a chunk, the engine can play that video segment as the answer rather than the whole video. Chunks that win the video answer slot are self-contained: they state a complete answer in a full sentence ('To clean suede shoes, brush the surface dry first…') rather than a fragment that only makes sense with the surrounding video ('So the first thing you wanna do here is…'). Writing a spoken script so each answer is a distinct, front-loaded, timestampable moment — and shipping a real punctuated transcript rather than relying on auto-captions — is how a creator makes their video chunk cleanly and become retrievable.